Information processing device and translation method

Abstract

To provide an information processing device, a translation method, and a program capable of reproducing and translating the use of language characteristic of a character into another language. [Solution] The manga translation system 10 extracts lines in a first language associated with a character profile indicating the character's attributes from manga frames, translates the extracted lines from the first language to a second language, and a correction unit corrects the lines translated into the second language so that the language used is in accordance with the associated character profile.

Description

[Technical Field]

[0001] The present invention relates to an information processing device and a translation method. [Background technology]

[0002] In recent years, so-called web manga, which are digitized manga that can be viewed on websites, have become widespread. In web manga, the dialogue is translated from Japanese into other languages ​​so that it can be viewed not only in Japan but also outside of Japan.

[0003] Patent Document 1 describes a method for obtaining manga information in which image information is associated with text information and audio information from a server, and performing control to display and output the text information and audio information corresponding to the same image information in different languages. Identification information for identifying characters appearing in the manga is added to the manga information, and a language combination is independently assigned to each character based on the identification information. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2014-007561 Summary of the Invention [Problem to be solved by the invention]

[0005] Here, characters appearing in manga may have unique language (phrases) that characterize the character. However, in Patent Document 1, although the language used for each character can be selected based on the identification information, the language used for each character is not reproduced and translated into another language. Furthermore, in order to translate the language used by each character into another language, a translator is required.

[0006] SUMMARY OF THE INVENTION It is therefore an object of the present invention to provide an information processing device and a translation method that can reproduce and translate the language characteristic of a character into another language. [Means for solving the problem]

[0007] The information processing device of the present invention includes an extraction unit that extracts lines in a first language associated with a character attribute from a manga frame, a translation unit that translates the lines extracted by the extraction unit from the first language into a second language, and a correction unit that corrects the lines translated into the second language by the translation unit so that the lines are worded in accordance with the associated attribute. Equipped with.

[0008] The translation method of the present invention includes a first step in which an extraction unit extracts lines in a first language associated with character attributes from manga frames; a second step in which a translation unit translates the lines extracted by the extraction unit from the first language into a second language; and a third step in which a correction unit corrects the lines translated into the second language by the translation unit so that the language used corresponds to the associated attributes. [Effects of the Invention]

[0009] According to the present invention, the language characteristic of a character can be reproduced and translated into other languages. [Brief explanation of the drawings]

[0010] [Figure 1] FIG. 1 is a configuration diagram of a manga translation system according to a first embodiment. [Figure 2] FIG. 2 is a functional block diagram of a server related to manga translation processing in the first embodiment. [Figure 3] 1 is a flowchart showing the flow of manga translation processing in the first embodiment. [Figure 4] FIG. 10 is a functional block diagram of a server related to manga translation processing in the second embodiment. [Figure 5] 10 is a flowchart showing the flow of manga translation processing in the second embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0011] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. Note that the embodiment described below shows an example of how the present invention can be implemented, and the present invention is not limited to the specific configuration described below. When implementing the present invention, a specific configuration corresponding to the embodiment may be appropriately adopted.

[0012] (First embodiment) Figure 1 is a configuration diagram of a manga translation system 10 according to this embodiment. The manga translation system 10 includes at least one personal computer (hereinafter referred to as "PC") 12 and a server 14. The PC 12 and the server 14 are communicatively connected via a network 16, which may include a wireless communication network or a wired communication network.

[0013] PC 12 is an information processing device such as a desktop computer, laptop computer, smartphone, or tablet device equipped with a display 13. PC 12 is operated by an operator who translates manga to be made available on a website. Specifically, the operator uses PC 12 to select the manga to be translated, and server 14 executes manga translation processing on the selected manga.

[0014] The server 14 is an information processing device that stores data on multiple manga works and performs manga translation processing on manga works selected by an operator. The manga works stored as data on the server 14 are, for example, image data obtained by scanning paper-based manga works, or manga works generated as digital data. In this embodiment, the manga works are formed from multiple frames and are divided into pages. Manga works are also specified by the number of volumes and chapters of the published works.

[0015] The manga translation process of this embodiment translates the lines of a manga character from a first language into a second language and corrects the lines translated into the second language so that the language is appropriate for the attributes associated with the character. This allows the manga translation process of this embodiment to reproduce and translate the language that characterizes the character into another language. Note that language refers to the unique expressions and verbal expressions that characterize the character.

[0016] In the following description, the first language is Japanese and the second language is English, but this is just an example and is not limited to this as long as the first and second languages ​​are different. The attributes associated with a character are called a character profile. A character profile is an attribute that characterizes the character's way of speaking, such as a young man, a gentleman, or an aggressive person. The character profile may not only be a typical attribute such as a young man, a gentleman, or an aggressive person, but may also be an attribute specific to a specific character that appears in a manga.

[0017] FIG. 2 is a functional block diagram relating to the manga translation process executed by the server 14.

[0018] The server 14 includes a storage unit 20, a communication unit 22, and a calculation unit 24.

[0019] The storage unit 20 stores programs for executing manga translation processing, manga data, and various other programs and data. In this embodiment, the storage unit 20 stores, for each character profile, translation correction data that indicates the wording of the second language corresponding to the character profile. The translation correction data is, for example, data that associates general wording in various second languages ​​with wording corresponding to the character profile.

[0020] The communication unit 22 transmits and receives data to and from the PC 12 and other information processing devices via the network 16 .

[0021] The calculation unit 24 is a computer such as a CPU (Central Processing Unit), and includes an extraction unit 30, a translation unit 32, a translation correction unit 34, a dialogue allocation unit 36, and a speech bubble processing unit 38. These functions are realized by executing a program stored in the storage unit 20 provided in the server 14.

[0022] The extraction unit 30 extracts lines in a first language associated with a character profile from a frame of the manga. The extraction unit 30, for example, performs image recognition on the manga to identify and extract manga characters and lines. The lines extracted from the manga are in a first language, for example, Japanese. Specifically, the extraction unit 30, for example, performs image recognition on speech bubbles and recognizes the characters contained in the speech bubbles as lines. Furthermore, the extraction unit 30, for example, identifies characters from the manga based on character data indicating characteristic features of the manga characters. The extraction unit 30 also extracts the position and size of the speech bubbles in which the lines are written.

[0023] The lines extracted by the extraction unit 30 are text data associated with, for example, the character who spoke the lines, the display position of the lines, etc. The display position of the lines is, for example, identification information such as the speech bubble in which the lines are displayed, the episode number in which the lines are displayed, the page in which the lines are displayed, the frame in which the lines are displayed, etc.

[0024] Furthermore, a character profile indicating the character who uttered the line is associated with the line extracted by the extraction unit 30. Therefore, a character profile is associated with each character appearing in a manga.

[0025] For example, if a character and dialogue are included in the same frame, the dialogue is considered to have been spoken by a character included in the same frame, and a character profile is associated with the dialogue. Alternatively, the character in the direction indicated by the speech bubble containing the dialogue may be identified, and the character profile of that character may be associated with the dialogue. Alternatively, if a frame contains only dialogue and no character, the dialogue may be considered to have been spoken by a character included in a frame before or after the frame, and a character profile may be associated with the dialogue. The method of associating a character profile with a dialogue is not limited to the above, and other methods may also be used. For example, a worker may manually associate a character profile for each line while checking the lines.

[0026] It is not necessary for all characters appearing in the manga to be associated with a character profile. Lines uttered by characters who are not associated with a character profile will not be associated with a character profile.

[0027] The translation unit 32 translates the lines extracted by the extraction unit 30 from the first language to the second language. The translation of the lines from the first language to the second language is a standard translation performed by, for example, machine translation, and is output as text data. A standard translation is a literal translation of the lines in the first language into the second language, and does not necessarily result in a translation that conforms to the character profile associated with the lines. Note that the tool (translation model) used for machine translation by the translation unit 32 is not limited.

[0028] The translation correction unit 34 corrects the lines translated into the second language by the translation unit 32 so that the language used conforms to the associated character profile. That is, in the manga translation process of this embodiment, the lines translated from the first language to the second language are corrected according to the character profile associated with the lines.

[0029] The translation correction unit 34 of this embodiment corrects the lines translated into the second language based on the translation correction data indicating the wording of the second language corresponding to the character profile stored in the storage unit 20.

[0030] Next, a specific example of dialogue translation and correction will be described.

[0031] For example, if the line in the first language, Japanese, is "Good morning. It's nice weather today, isn't it? What are your plans for today?", the translation unit 32 translates it into the second language, English, as "Good morning. The weather is nice today, isn't it? What are your plans for today?"

[0032] On the other hand, if the character profile of the line is a young person, the translation correction unit 34 corrects the line to "Morning! Nice weather, huh? Got any plans for today?" using translation correction data with the character profile set to "young person." Also, for example, if the character profile of the line is a gentleman, the translation correction unit 34 corrects the line to "Good morning. The weather is quite delightful today, wouldn't you agree? Might I inquire about your plans for the day?" using translation correction data with the character profile set to "gentleman." Also, for example, if the character profile of the line is an aggressive person, the translation correction unit 34 corrects the line to "Morning. The weather's decent, I guess. Got anything actually productive planned for once?" using translation correction data with the character profile set to "aggressive person."

[0033] The dialogue assignment unit 36 ​​assigns the dialogue translated into the second language to the speech bubble of the frame associated with the dialogue, so that the dialogue translated into the second language is displayed in the speech bubble of the manga.

[0034] The speech bubble processor 38 enlarges the speech bubble when the translated lines into the second language do not fit in the speech bubble that displayed the lines before translation into the second language. For example, when translating from Japanese to English, the translated lines into English may be longer than the Japanese lines. In such a case, the speech bubble that displayed the Japanese lines may be too small for the English lines to fit in. In such a case, the speech bubble processor 38 enlarges the speech bubble, and the line assignment unit 36 ​​assigns the translated lines to the enlarged speech bubble.

[0035] Note that whether or not the speech bubble needs to be enlarged is determined by speech bubble processor 38, for example, based on the number of characters in the first language and the number of characters in the second language. For example, the number of characters in the dialogue before and after translation is compared, with two half-width alphabetic characters being considered as one Japanese character, and an enlargement rate for the speech bubble is calculated based on the difference in the number of characters before and after translation. If the enlargement rate exceeds 1, it is determined that the speech bubble needs to be enlarged. Then, the speech bubble is enlarged according to the enlargement rate, and the translated dialogue is assigned to the enlarged speech bubble.

[0036] FIG. 3 is a flowchart showing the flow of the manga translation process of this embodiment.

[0037] First, in step 100, the extraction unit 30 extracts lines from the manga for each character and converts them into text data. The extracted lines are associated with a character profile, but not all lines need to be associated with a character profile. For example, a character who is not a main character, or who does not use distinctive language in their lines, will not be associated with a character profile.

[0038] In the next step 102, the translation unit 32 translates the extracted lines in the first language into the second language.

[0039] In the next step 104, the translation correction unit 34 determines whether or not a character profile is associated with the lines translated into the second language, and if the determination is affirmative, the process proceeds to step 106, and if the determination is negative, the process proceeds to step 110. In other words, if a character profile is not associated with the lines, no correction is made by the translation correction unit 34.

[0040] In step 106, the translation correction unit 34 corrects the lines translated into the second language according to the character profile associated with the lines using the translation correction data stored in the storage unit 20, and the process proceeds to step 108.

[0041] In step 108, speech bubble processor 38 calculates the speech bubble enlargement rate to determine whether or not the speech bubble needs to be enlarged.

[0042] In the next step 110, it is determined whether or not the translation and enlargement rate calculation processes have been completed for all lines. If the determination is affirmative, the process proceeds to step 112, whereas if the determination is negative, the process returns to step .

[0043] In step 112, the line translated into the second language is assigned to the speech bubble corresponding to the line by the line assignment unit 36. At this time, if the expansion factor exceeds 1 in step 108, the translated line is assigned to the speech bubble expanded by that expansion factor.

[0044] The lines translated by the manga translation process are output to the PC 12 and checked for appropriateness by a translator. The translator may be a different person from the operator. If the lines in the second language corrected according to the character profile are inappropriate, they are corrected to appropriate lines, and translation correction data is corrected or added.

[0045] Note that lines whose translations have been corrected may be highlighted to make them easier for the translator to check. For example, highlighting may be done by highlighting lines whose translations have been corrected in a brighter color to make them stand out. Furthermore, to make it easier to check the appropriateness of lines corrected by the translation correction unit 34, the manga translation system 10 may output three lines that can be compared: the lines in the first language, the lines translated into the second language, and the lines in the second language corrected according to the character profile. Note that outputting means, for example, displaying them on the display 13 of the PC 12 or making them printable on a printer.

[0046] Furthermore, there may be cases where the translation of lines into the second language by the translation unit 32 is inappropriate. In such cases, the translator may correct the lines translated into the second language, and the corrected lines may be corrected by the translation correction unit 34. The corrected lines in the second language may be associated with the lines in the first language before translation, and may be used as training data for the translation model used by the translation unit 32.

[0047] As explained above, the manga translation process of this embodiment corrects the lines translated from a first language to a second language so that the language used corresponds to the character profile, so that the language characteristic of the character can be reproduced and translated into another language.

[0048] (Second embodiment) The manga translation process of this embodiment uses a machine-learned correction model to correct dialogue translated from a first language to a second language so that the language usage matches the character profile. Note that the configuration of the manga translation system 10 of the second embodiment is the same as that of the manga translation system 10 of the first embodiment, so a description thereof will be omitted.

[0049] 3 is a functional block diagram of the manga translation process executed by the server 14 of this embodiment. Note that the same components in the functional block diagram of the second embodiment as those in the first embodiment are designated by the same reference numerals, and their description will be omitted.

[0050] The calculation unit 24 includes a model generation unit 50, an extraction unit 30, a translation unit 32, a translation correction unit 52, a dialogue assignment unit 36, and a speech bubble processing unit 38.

[0051] The model generation unit 50 generates a correction model through machine learning using lines in a second language that are the language of the character indicated by the character profile as training data. The correction model receives lines that are associated with the character profile and translated into the second language, and outputs lines that have been corrected to the language that corresponds to the character profile. The correction model is stored in the storage unit 20.

[0052] The training data for the correction model of this embodiment is data that associates, for example, lines expressed in general language usage in the second language, lines in the second language that have the same meaning as the lines but are expressed in language that corresponds to the character profile, and a character profile.

[0053] The training data for the correction model may be lines that have been corrected by the correction model and then corrected by a translator. This allows more appropriate corrected lines to be used as training data, enabling the model generation unit 50 to generate a more accurate correction model.

[0054] Furthermore, a different correction model may be generated for each character profile. For example, a correction model corresponding to a character profile of a young person and a correction model corresponding to a character profile of a gentleman may be generated.

[0055] The translation correction unit 52 corrects the inputted lines to a wording that corresponds to the character profile by inputting the text data of the lines translated into the second language by the translation unit 32 into the correction model generated by the model generation unit 50, and outputs the corrected lines. As described above, when the correction model differs for each character profile, the translation correction unit 52 corrects the lines using the correction model that corresponds to the character profile associated with the lines.

[0056] Furthermore, scene information indicating the situation, context, etc. of a manga scene may be associated with the dialogue and used as training data for training the correction model. For example, the scene information of the dialogue to be corrected is estimated by analyzing the time series of text data of the dialogue to be corrected (such as the dialogue before and after). The scene information may include, for example, the emotions of the character, such as joy, anger, sadness, or happiness, and the situation around the character who spoke the dialogue. Note that the scene information may also be manually associated with the dialogue by an operator. The scene information may also be used in a translation model for translating dialogue from a first language to a second language.

[0057] 5 is a flowchart showing the flow of the manga translation process of this embodiment. Steps in the manga translation process of the second embodiment that are the same as those in the first embodiment are given the same reference numerals, and their explanation will be omitted.

[0058] If the determination in step 104 is affirmative, the process proceeds to step 200.

[0059] In step 200, the lines translated into the second language are input to the translation correction unit 52, and the correction model corrects the input lines to the language used in accordance with the character profile and outputs them, and the process proceeds to step 108.

[0060] As described above, the manga translation process of this embodiment corrects the lines translated into the second language to the language used by the character indicated in the character profile by inputting the lines translated into the second language into a correction model that has been machine-learned using training data from the lines in the second language.

[0061] Although the present invention has been described above using the above-mentioned embodiment, the technical scope of the present invention is not limited to the scope described in the above-mentioned embodiment. Various changes or improvements can be made to the above-mentioned embodiment without departing from the gist of the invention, and such changes or improvements are also included in the technical scope of the present invention.

[0062] In the above embodiment, the form in which dialogue is extracted from a manga and converted into text data has been described, but the present invention is not limited to this. Instead of extracting dialogue from a manga, the dialogue may be created as text data in advance and a character profile may be associated with the dialogue.

[0063] The following additional notes are provided regarding the above-described embodiment.

[0064] [Appendix 1] The information processing device of Supplementary Note 1 includes an extraction unit that extracts lines in a first language associated with a character attribute from a manga frame; a translation unit that translates the lines extracted by the extraction unit from the first language into a second language; and a correction unit that corrects the lines translated into the second language by the translation unit so that the lines are expressed in a language that corresponds to the associated attribute. Equipped with.

[0065] According to this configuration, the lines translated from the first language to the second language are corrected to use language that corresponds to the attributes associated with the character, so that the language that is characteristic of the character can be reproduced and translated into another language.

[0066] [Appendix 2] The information processing device of Appendix 1 may further include a memory unit that stores, for each attribute, the wording of the second language corresponding to the attribute, and the correction unit may correct the lines translated into the second language based on translation correction data that indicates the wording of the second language that corresponds to the attribute stored in the memory unit.

[0067] According to this configuration, the language used in the second language according to the attributes of the character is stored in a memory unit for each attribute, and the translated lines are corrected based on the stored language, so that the language characteristic of the character can be reproduced in another language and easily translated.

[0068] [Appendix 3] In the information processing device of Appendix 1, the correction unit may input lines translated into the second language and associated with the attributes into a correction model that has been machine-learned using training data of lines in the second language that are the language of the character indicated by the attributes, and output the lines corrected to language that corresponds to the attributes.

[0069] According to this configuration, the lines translated into the second language are corrected to use language that corresponds to the character's attributes based on a machine-learned correction model, so that the character's characteristic language can be reproduced in other languages ​​and easily translated.

[0070] [Appendix 4] The information processing device of Supplementary Note 3 may use the lines corrected by the correction model and then corrected by a translator as training data for the correction model.

[0071] According to this configuration, a more accurate correction model can be generated.

[0072] [Appendix 5] In the information processing device of any one of Supplementary Note 1 to Supplementary Note 4, if the attribute is not associated with the line, the correction unit may not perform the correction.

[0073] According to this configuration, unnecessary correction can be prevented.

[0074] [Appendix 6] An information processing device according to any one of Supplementary Notes 1 to 5 may enlarge the speech bubble when the dialogue translated into the second language does not fit in the speech bubble in which the dialogue before translation into the second language was displayed.

[0075] According to this configuration, even when the dialogue is translated into a second language, it is possible to prevent the visibility of the dialogue from becoming poor.

[0076] [Appendix 7] The translation method of Appendix 7 includes a first step in which an extraction unit extracts lines in a first language associated with a character attribute from a manga frame, a second step in which a translation unit translates the lines extracted by the extraction unit from the first language into a second language, and a third step in which a correction unit corrects the lines translated into the second language by the translation unit so that the lines are worded in accordance with the associated attribute.

[0077] According to this configuration, the lines translated from the first language to the second language are corrected to use language that corresponds to the attributes associated with the character, so that the language that is characteristic of the character can be reproduced and translated into another language. [Explanation of symbols]

[0078] 10 Manga Translation System 14 Server (information processing device) 20 Memory section 30 Extraction part 32 Translation Department 34 Translation Correction Unit (Correction Unit) 52 Translation Correction Unit (Correction Unit)

Claims

1. an extraction unit that extracts lines in the first language associated with character attributes from manga frames; a translation unit that translates the lines extracted by the extraction unit from the first language into a second language; a correction unit that corrects the lines translated into the second language by the translation unit so that the lines are expressed in a manner that corresponds to the associated attribute; An information processing device comprising:

2. a storage unit that stores, for each attribute, the wording of the second language corresponding to the attribute; Equipped with the correction unit corrects the lines translated into the second language based on translation correction data indicating wording in the second language corresponding to the attribute stored in the storage unit. The information processing device according to claim 1 .

3. 2. The information processing device according to claim 1, wherein the correction unit inputs lines translated into the second language and associated with the attributes into a correction model that has been machine-learned using training data of lines in the second language that are used by the character indicated by the attributes, and outputs the lines corrected to language that corresponds to the attributes.

4. The information processing device according to claim 3 , wherein the lines corrected by the correction model and then corrected by a translator are used as training data for the correction model.

5. The information processing device according to claim 1 , wherein if the attribute is not associated with the line, the correction unit does not perform correction.

6. The information processing device according to claim 1 , wherein the speech bubble is enlarged when the translated lines into the second language do not fit in a speech bubble that displayed the lines before translation into the second language.

7. A first step in which an extraction unit extracts lines in a first language associated with character attributes from manga frames; a second step of translating the lines extracted by the extraction unit from the first language into a second language by a translation unit; a third step in which a correction unit corrects the lines translated into the second language by the translation unit so that the lines are expressed in a manner that corresponds to the associated attribute; A translation method performed by an information processing device.

Images